Coaxial cable television systems have been in widespread use for many years and extensive networks have been developed. The extensive and complex networks are often difficult for a cable operator to manage and monitor. A typical cable network generally contains a headend which is usually connected to several nodes which provide content to a cable modem termination system (CMTS) containing several receivers, each receiver connects to several network elements of many subscribers, e.g., a single receiver may be connected to hundreds of network elements. In many instances several nodes may serve a particular area of a town or city. The network elements communicate to the CMTS via upstream communications on a dedicated band of frequency.
Cable networks are also increasingly carrying signals which require a high quality and reliability of service, such as voice communications or Voice over IP (VoIP) communications. Any disruption of voice or data traffic is a great inconvenience and often unacceptable to a subscriber. Various factors may affect the quality of service, including the quality of the upstream channels. Cable networks use various management techniques to control their networks, including return path spectrum management techniques. Return path spectrum management generally allows a cable network operator to reassign network elements to different communication parameters or different communication devices, e.g. receivers, to optimize use of the bandwidth spectrum available under certain conditions.
Traditional return-path spectrum management techniques in the cable industry generally revolve around three classical approaches: 1) fast Fourier transform (FFT) or traditional return-path noise power measurements, 2) Packet or Bit error rate tests, and 3) hop-and-hope or trial and error approaches. Clearly the third approach is a poor solution. The second approach requires longer measurement times due to the long period required to collect a statically significant numbers of packets/bits on which to base a metric and in addition, is subject to the random nature of which network elements are sending data (for example, poor performers might bias the results). In addition, the second approach requires the network to make errors prior to triggering network changes. For applications which require error-free or near error-free communications, this is an unacceptable approach. The first approach is solely based upon noise power measurements and requires that the channel be deactivated while the actual noise power measurement is made. A better approach is one which does not impact active data services, considers both noise and distortion affects, and indicates when signal quality is marginal but not yet making actual errors. The modulation error ratio (MER) is often considered the perfect measurement. However, MER for a single channel can vary a great deal (4 to 12 dB) depending upon which cable modem (network element) is selected for the measurement. Further, transient conditions can impact single measurements and must be understood in order to make accurate spectrum management decisions.
Previous techniques which focus on packet/bit error rates or traditional power based Modulation Error Ratio (MER) measurements generally use a composite (average) MER to evaluate the currently active channel and base modulation agility decisions. This approach, however, contains two significant weaknesses: 1) the composite MER reflects the “currently active” cable network elements as measured by data being passed on the upstream, and 2) the composite MER is an average which means that (generally) half the network elements are exhibiting MERs which are lower while many are exhibiting MERs which are higher.
Depending upon what the variance of MER is across the network element population, the range of actual MER performance of the network element population may be wide or narrow. When utilizing the average MER to monitor modulation changes, a narrow distribution of MER values among a plurality of network elements (meaning all network elements are yielding very similar MER readings) are good and allow the network operator to make modulation changes in which all network elements may continue to pass data. However, if the distribution is wide, and the network operator bases the modulation configuration changes on the average, there is a significant risk that some of the network elements (which are operating at MERs much lower than the channel average) will no longer be able to pass data on such a channel. For example, typical plant conditions are expected to yield a spread of between 4 and 12 dB. With 3 dB differences required to support each modulation type (QPSK, 8QAM, 16QAM, 32QAM, 64QAM, 128QAM, 256QAM), this implies that network elements are typically distributed across 2 to 4 different supported modulation types.
Generally, MER is a good predictor of what level of modulation (QPSK, 8QAM, 16QAM, 32QAM, 64QAM, 128QAM, or 256QAM) that a communications channel can support. With most DOCSIS 2.0 CMTS's deployed today, the operator can retrieve each cable modem's or MTA's MER and could therefore generally predict whether the channel could be reconfigured to run a higher modulation and therefore achieve a higher bit rate (throughput) on the channel. The one major exception to the prediction is the impacts of non-linearities (specifically such factors as overstressing the optical link—either laser or optical receiver, and in some remote cases, stressing an RF amplifier). Non-linearities are not reflected necessarily reflected at lower level modulations. Therefore, the user must increase the modulation level while simultaneously monitoring the MER measurement. If a significant change in the MER is experienced as the modulation level is increased, then it may be correctly inferred that a non-linearity is present and higher modulations require further analysis. Thus, the user must perform measurements at higher modulations (risking impacts to data communications) to successfully predict these modulation configuration changes. The user clearly does not want to change the modulations of all his modems/MTAs while providing data services in order to perform this characterization. What is needed is the ability to selectively change the modulation level of a single modem/MTA device without impacting all of the others.